Reviewing electricity production cost assessments

MPRA Munich Personal RePEc Archive Reviewing electricity production cost assessments Simon Larsson and Dean Fantazzini and Simon Davidsson and Sven Kullander and Mikael Hook Global Energy Systems, Department of Earth Sciences, Uppsala University (Sweden), Moscow School of Economics, Moscow State University (Russia), Global Energy Systems, Department of Earth Sciences, Uppsala University (Sweden), Royal Swedish Academy of Sciences, Box 50005, SE-104 05 Stockholm, Sweden, Global Energy Systems, Department of Earth Sciences, Uppsala University (Sweden) September 2013 Online at http://mpra.ub.uni-muenchen.de/50306/ MPRA Paper No. 50306, posted 1. October 2013 12:29 UTC

Accepted by Renewable & Sustainable Energy Reviews http://www.journals.elsevier.com/renewable-and-sustainable-energy-reviews/ Reviewing electricity production cost assessments Simon Larsson 1,2, Dean Fantazzini 3, Simon Davidsson 1, Sven Kullander 2, Mikael Höök 1, 1 Global Energy Systems, Department of Earth Sciences, Uppsala University, Villavägen 16, SE-752 36, Uppsala, Sweden. Web: http://www.geo.uu.se 2 Royal Swedish Academy of Sciences, Box 50005, SE-104 05 Stockholm, Sweden. Web: http://www.kva.se 3 Moscow School of Economics, Moscow State University, 1, Building 61, Leninskie Gory 119992, Moscow, Russia. Web: www.mse-msu.ru Abstract A thorough review of twelve recent studies of production costs from different power generating technologies was conducted and a wide range in cost estimates was found. The reviewed studies show differences in their methodologies and assumptions, making the stated cost figures not directly comparable and unsuitable to be generalized to represent the costs for entire technologies. Moreover, current levelized costs of electricity methodologies focus only on the producer's costs, while additional costs viewed from a consumer perspective and on external costs with impact on society should be included if these results are to be used for planning. Although this type of electricity production cost assessments can be useful, the habit of generalizing electricity production cost figures for entire technologies is problematic. Cost escalations tend to occur rapidly with time, the impact of economies of scale is significant, costs are in many cases site-specific, and country-specific circumstances affect production costs. Assumptions on the cost-influencing factors such as discount rates, fuel prices and heat credits fluctuate considerably and have a significant impact on production cost results. Electricity production costs assessments similar to the studies reviewed in this work disregard many important cost factors, making them inadequate for decision and policy making, and should only be used to provide rough ballpark estimates with respect to a given system boundary. Caution when using electricity production cost estimates are recommended, and further studies investigating cost under different circumstances, both for producers and society as a whole are called for. Also, policy makers should be aware of the potentially widely different results coming from electricity production cost estimates under different assumptions. Keywords: Electricity generation costs, levelized costs, energy market analysis, cost comparisons, policy implications Corresponding author: Mikael Höök, Mikael.Hook@geo.uu.se 1

Nomenclature CASES CC CCGT CHP CHP-B CHP-W DECC DR EIA EU EUR EUSUSTEL EWEA GBP GDP HC HP IDC IEA IHS IMF kw e MIT MW e MWh e NEA NEEDS NP OCGT PB PV PwC PWR R&D SKGS USD VGB WP Cost Assessments for Sustainable Energy Systems Coal Condensing Combined Cycle Gas Turbine Combined Heat and Power Combined Heat and Power Biomass Combined Heat and Power Waste Department of Energy and Climate Change Discount Rate U.S. Energy Information Administration European Union Euro European Sustainable Electricity European Wind Energy Association British Pound Gross Domestic Product Heat Credit Hydropower Interest During Construction International Energy Agency Information Handling Services International Monetary Fund kilowatt electrical Massachusetts Institute of Technology MegaWatt electrical Megawatt-Hour electrical Nuclear Energy Agency New Energy Externalities Development for Sustainability Nuclear Power Open Cycle Gas Turbine Parsons Brinckerhoff Photovoltaics PricewaterhouseCoopers Pressurized Water Reactor Research and Development Skogen, Kemin, Gruvorna och Stålet (Swedish) United States Dollar VGB PowerTech e.v. Wind power 2

1. Introduction The last decade has witnessed an increased number of studies examining electricity generation or production costs 1 from alternative power generating technologies, and references to such studies can often be heard in the public debate. In the European Union (EU), electricity production costs assessments have been used to justify various forms of economic support schemes for renewable power generating technologies [1,2], and to estimate necessary levels of CO 2 emission fees, such as those used within the EU Emission Trading System, needed to narrow the gap between fossil fuelled technologies and other low-carbon alternatives [3]. Recent literature in this field includes the works by NEEDS [4], IEA and NEA [5], EIA [6], VGB [7], Elforsk [8], and DECC [9]. It is common to find the cost figures reported in these studies to vary greatly. This is problematic since electricity generation cost assessments are typically used to rank different power generating technologies based on the expected generation cost and to estimate economic subsidies or penalty charges needed. As such, generation cost assessments are important in energy decision and policy making. This paper focuses on costs for producers, i.e. costs that that can be directly attributed to the investment and operation of power plants. Such costs will be named production costs of electricity in this paper. However, power plants also cause socio-environmental damages during their construction, operation and dismantling, for instance, due to the emission of environmentally unfriendly and hazardous gases, particles and greenhouse gases. Such costs are called external costs, since they are costs not incurred by the power plant owners 2. External costs are discussed in this paper, but in a more qualitative way. Whenever they are discussed and/or added to production costs, the wording electricity generation costs will be used. This paper 3 reviews several electricity production cost assessments with the purpose to identify the differences in the methodologies and in the underlying assumptions that can explain the often diverging cost figures reported in the literature. These differences are discussed with special attention to some major factors influencing the stated electricity production costs. Moreover, issues regarding the current methodologies are also highlighted and discussed from a perspective of energy policy making. A thorough analysis of twelve studies of electricity production costs published in the year 2007 or later, is carried out. Some of these studies were conducted or commissioned by government departments [10 12]; others were performed by energy agencies [5], by research and development companies [8], consulting firms [13], and by European Commission funded projects [4, 14, 15]. The study by Förnybart.nu [16] is a compilation of different studies [17 20 and was also analyzed in this study. The recent review by Branker et al. [21] about solar photovoltaic levelized cost of electricity production was also included. The review about low-carbon baseload generating technologies by Nicholson et al. [22] was not considered since the electricity production costs presented appear too generalized and it was not deemed possible to distinguish different cost components (for example, external costs are not separated from electricity production costs and the discount rates used are not presented). Freeman [23] provided an overview of the environmental costs of electricity, while Damen et al. [24] compared electricity and hydrogen production systems with CO2 capture and storage, but none of them are included since they appear quite outdated. Finally, 1 A more suitable name would be electrical energy production or generation costs. The use of electricity as an equivalent of electrical energy is commonly accepted and it will therefore be retained in this paper. 2 See the report by the European Commission [28] for more details. 3 This paper is based on the work conducted by Larsson [29] on behalf of the Energy Committee at the Royal Swedish Academy of Sciences. For details and further information related to the topics discussed in this paper, please see Larsson [29]. 3

we remark that Khatib [25] provided an overview of the joint report by the IEA and the OECD Nuclear Energy Agency [5], which is one of the 12 studies included in our review. The paper is structured as follows: section 2 presents an overview of the electricity production cost estimates from different studies. Section 3 highlights some factors that have a major influence on the stated costs, and which can explain the wide range of published estimates. Section 4 is devoted to the implications of using these electricity production cost estimates for policy making. Finally, section 5 provides a concluding discussion of the main findings. 4

2. Overview of current electricity production cost estimates The main focus of this study is on current cost figures of commercial power generating technologies: to avoid outdated cost figures, no study published earlier than the year 2007 was included in this review (see Table 1). Table 1. Reviewed studies and investigated power generating technologies. Consulting firms ARUP [10] and Parsons Brinckerhoff [12] conducted the latest U.K. Department of Energy and Climate Change (DECC) studies on electricity production costs. The SKGS study (a cooperation between Swedish industry associations) on electricity production costs was performed by the professional services firm PricewaterhouseCoopers (PwC) [13]. All abbreviations are listed in the nomenclature. Some studies have more than one year in the references since they were divided in several different papers and/or because the information is located at several different places. Study Included power generating technologies CC CHP-B CHP-W HP CCGT CCGT-CHP OCGT PV NP WP on WP off CASES [14] x x x x X x x x x x Elforsk [8] x x x x x X x x x x EUSUSTEL [15] x x x X x x x x Förnybart.nu [16] x x x x x IEA & NEA [5] x x X x x x x NEEDS [4] x x x x x x x SKGS [13] x x x DECC [9] x x x x x X x x x x x EWEA [18] x x EIA [11] x x x x x x x x VGB [7] x x x x x x x x Branker et al. [21] x External costs, due to the potential impact of power plants on, for example, human health, ecosystems, crops, buildings and climate change are not considered. Costs induced by taxes and subsidies are also excluded, despite the fact that such costs can be significant, since they vary among different countries and are decided by governments. However, external costs, subsidies, taxes and differences in technical characteristics have to be considered in decision and policy making. The electricity system is simply affected in different ways depending on the technology with some power plants in more need of balancing power, back-up capacity and transmission investments than others. Therefore, these will be discussed in Section 4 which considers the policy implications of assessments for electricity generation costs. Estimates of future electricity production costs that might change in the future due to, for example, learning curves are also outside the scope of this review. From an investor s perspective, the future assumptions about the income coming from produced electricity are crucial for investment decisions. In liberalized electricity markets, investors face greater risk as future electricity price levels are uncertain [26]. Moreover, current electricity production cost methodologies generally fail to account for the fact that dispatchable power generating technologies have an economic and technical advantage over non-dispatchable alternatives, since the latter only produce electrical energy during favorable weather conditions. However, our assessments attempts to only consider studies of production costs. The only exceptions to this rule are represented by some studies which incorporate the costs of carbon dioxide emission fees and backup costs for intermittent power generating 5

technologies into the normal electricity production costs. Given that these costs were not presented as external cost factors or taxes, we opted to keep them to avoid complicating the comparison with the other reviewed studies (see Sections 3.6-3.7). Costs are presented in United States Dollar (USD). Exchange rates used for conversions from other currencies are presented in Table 2. Cost figures originating in foregoing years were recalculated to 2011 values by using the (country specific) GDP deflators provided by the IMF [27]. Note that many of the reviewed studies use other currencies than USD, so that the presented cost figures could be sensitive to changes in exchange rates. Table 2. Assumed exchange rates when original costs were not presented in USD. Current data obtained from Federal Reserve. Currency Exchange rates used EUR 1.393 USD/EUR GBP 1.604 USD/GBP SEK 0.154 USD/SEK Levelized costs of electricity production for non-renewable power generating technologies are shown in Figure 1, wind power and combined heat and power biomass-fired plants in Figure 2 and hydropower and photovoltaics in Figure 3. USD/MWhe 180 170 160 150 140 130 120 110 100 90 80 70 60 50 40 30 20 10 0 upper quartile box lower quartile box mean Figure 1. Levelized costs of electricity production for non-renewable power generating technologies. The cost figures are visualized in box plots showing the minimum value, lower quartile, median, upper quartile, and the highest value. The mean value is also included in the figure. Costs induced by assumptions on carbon dioxide emission fees (section 3.6) are not included. Studies that do not present electricity production costs by using a 5% or a 10% discount rate are not included in this figure (i.e. [5], [7]; [8], [12], [14], [15], and [16]). 6

USD/MWhe 360 340 320 300 280 260 240 220 200 180 160 140 120 100 80 60 40 20 0 Wind power onshore 5% DR upper quartile box lower quartile box mean Wind power onshore 10% DR Wind power offshore 5% DR Wind power offshore 10% DR CHP-B 5% DR CHP-B + heat credit 5% DR CHP-B 10% DR CHP-B + heat credit 10% DR Figure 2. Levelized costs of electricity production for wind power and combined heat and power biomass-fired plants. The cost figures are visualized in box plots showing the minimum value, lower quartile, median, upper quartile, and the highest value. The mean value is also included in the figure. Costs induced by assumptions on backup costs for intermittent power generating technologies (section 3.7) are not included. Studies that do not present electricity production costs by using a 5% or a 10% discount rate are not included in this figure (i.e. [4], [5], [7], [8], [10], [14], and [15]). USD/MWhe 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 100 0 Reservoir hydro 5% DR upper quartile box lower quartile box mean Reservoir ROR hydro 5% hydro 10% DR DR Figure 3. Levelized costs of electricity production from hydropower and photovoltaics. The cost figures are visualized in box plots showing the minimum value, lower quartile, median, upper quartile, and the highest value. The mean value is also included in the figure. Costs induced by assumptions on backup costs for intermittent power generating technologies are not included. Studies that do not present electricity production costs by using a 5% or a 10% discount rate are not included in this figure (i.e. [4], [5], [7], [8], [10], [14], [15], and [20]). 7 ROR hydro 10% DR PV 5% DR PV 10% DR

Figures 1-3 show that the estimated producer costs vary quite considerably between different studies. Moreover, the dispersion of the estimates for hydropower and photovoltaics is much higher than that of the other technologies. This is due to a few extreme hydropower cost figures, together with the fact that electricity production costs for hydropower and photovoltaics tend to be highly site-specific. In this context it should be mentioned that the stated production costs are neither directly comparable with each other, nor can they be generalized to represent a general electricity production cost for a specific power-generating technology, due to the methodological differences and the varying assumptions present in the reviewed studies. As can be seen in the previous figures, the stated electricity production costs are sensitive to changes in the discount rate (DR), i.e. the interest rate used to determine the present value of future cash flows. Power generating technologies associated with high upfront capital costs are very sensitive to changes in discount rate, as it reflects the value put on time preferences. High upfront costs have to be valued against a stream of discounted future incomes. For example, a payment of 1000 $ occurring 10 years from now is worth 613 $ today at a discount rate of 5%, and 386 $ at a discount rate of 10%. As such, the value that is chosen for the discount rate can weight the decision towards one option or another. Moreover, electricity production costs from combined heat and power (CHP) plants are strongly influenced by assumptions on heat credits (HC), i.e. the value of the heat that is produced jointly with electrical energy. Assumptions on discount rates and the methodology used to calculate heat credits vary significantly across the studies, thereby contributing to the large dispersion of costs. The diverging assumptions about the fuel price are the main reason for the different cost estimates for CHP-B, CHP-W, CCGT-CHP, and lignite-fuelled power plant production costs. For most power generating technologies (with the exception of fossil-fuelled alternatives), the assumptions on investment costs differ significantly and contribute to additional variability in the final estimates. Another reason for discrepancies across studies is the assumption about the mode of operation for a power plant. For example, dispatchable power plants are in some studies assumed to be used in a base load mode, in other studies in intermediate or in peak load modes. Electricity productions costs are also dependent on the location of the power plant. This is visualized by Figures 4-5 which show that electricity production costs indeed seem to vary between different world regions. Differences in e.g. investment costs, fuel prices and labor costs results in different preconditions and thereby different electricity generation costs for different world regions [5], as Figures 4-5 indicate. Table 3 presents the reviewed studies geographical coverage (location) and their method of computation. Production costs presented in Elforsk [8] and other reviewed studies also show a strong influence of economies of scale, meaning that production costs are largely determined by the installed capacity, i.e. plant size. The cost-influencing factors mentioned are just examples of why studies differ. Some factors will be more thoroughly discussed in section 3, while others will not be dealt with mainly due to lack of transparency (i.e. lack of sufficient information, lumped cost figures, unknown impacts, etc.) in the reviewed studies. Production costs for waste-fuelled combined heat and power plants (CHP-W) are not shown in Figures 1-3. The reason is that such production costs can in fact be negative, since these power plants are usually paid for accepting waste for burning. This makes it difficult to exhibit CHP-W production costs together with production costs from other power-generating alternatives. Interested readers are referred to Larsson [29] for more details. 8

USD/MWh e 5% DR 0 20 40 60 80 100 120 140 160 180 N.America Europe Asia Nuclear Coal Gas Onshore Wind Nuclear Coal Gas Onshore Wind Nuclear Coal Gas Onshore Wind Figure 4. Comparison of levelized costs of electricity production in different world regions at 5 % discount rate. Only a small subset of studies provided regional data (i.e. [4], [5], [8], [14], and [15]). USD/MWh e 10% DR 0 20 40 60 80 100 120 140 160 180 200 220 240 260 N.America Europe Asia Nuclear Coal Gas Onshore Wind Nuclear Coal Gas Onshore Wind Nuclear Coal Gas Onshore Wind Figure 5. Comparison of levelized costs of electricity production in different world regions at 10 % discount rate. Only a smaller subset of the reviewed studied provided regional data (i.e. [5], [7], [8], [10], [12], [14], and [15]). 9

Table 3. Geographical coverage and method of computation. Study Geographic location Method of computation CASES [14] Europe Average Lifetime Levelised Generating Cost Elforsk [8] Sweden (Europe) Equivalent Annual Cost EUSUSTEL [15] Europe Average Lifetime Levelised Generating Cost Förnybart.nu [16] Sweden (Global) Literature review IEA & NEA [5] Europe (Global) Levelised Costs of Electricity NEEDS [4] Europe Average Lifetime Levelised Generating Cost SKGS [13] Sweden (Europe) Free Cash Flow DECC [9] United Kingdom Levelised Cost of Generation EWEA [18] Europe Levelised Cost EIA [11] United States Levelized Cost VGB [7] Europe Levelised Costs of Electricity Branker et al. [21] Global Levelized Cost of Electricity 3. Main reasons for diverging electricity production cost estimates 3.1 Methodological differences in the computation of the levelized cost of electricity The cost of generating electricity can be calculated in various ways. A widely accepted approach is the so called levelized cost of electricity (LCOE), or similar names such as average lifetime levelised generation cost (ALLGC), and levelised cost of generation (LCG). IEA & NEA [5], DECC [9], CASES [14], NEEDS [4] and EUSUSTEL [15] all used definitions of levelized cost of electricity similar to the formula presented below in Equation 1. Despite this, differences in methodology exist, and they are in some cases substantial enough to cause significant differences in results between studies. LCOE = t (investment t + O&M t + Fuel t + Others t ) (1 + d) t ) t (Electricity t (1 + d) t ) (1) Equation 1. Levelized cost of electricity formula. Expenditures for carbon dioxide emissions, decommissioning, refurbishment, etc. are captured in the parameter Others. Which assumptions are included as well as what they are, vary from study to study. A derivation of Equation 1 can be located in [5]. Observe that the carbon costs and the decommissioning costs mentioned in [5] is lumped under the Others category in Equation 1. Investment t investment expenditure in year t O&M t operation and maintenance expenditure in year t Fuel t fuel expenditure in year t Others t other expenditures in year t Electricity t electrical energy generation in year t d discount rate Equation 1 simply follows from the basic assumption that the present value of all discounted power-plant revenues has to be equal to the present value of all discounted powerplant costs. The LCOE is thereby equal to the constant average price at which electricity must be sold to reach breakeven over the economic lifetime of the power plant project 4. External 4 We remark that it is not electricity which is discounted in the denominator in Equation 1, but rather the value of the output that is produced. 10

costs, taxes and other costs not directly connected with a power plant (as part of its construction or operation), might also be included in the LCOE. However, these costs were not considered in the present study (see the previous discussion in section 2). A completely different valuation approach compared with the standard formula-based approaches to calculate the LCOE is found in the analysis by EIA [11]. In their analysis, production costs are calculated using an energy-economic model, whose results cannot be exactly recreated outside the model [30]. Even though the production costs are presented as levelized costs, the underlying approach is quite different from the formula-based methodologies such as the LCOE (Equation 1). Elforsk [8] and VGB [7] used the equivalent annual-cost method to calculate the investment-costs share of the total production cost, while cost components such as operation and maintenance (O&M) and fuel costs are assumed to be constant over time. In other cases production costs are calculated with confidential free cash flow models [13] or more or less unclear approaches [17, 19]. The expression levelized costs can be confusing, since such a commonly used terminology might imply that production costs are calculated using the same methodology, while this is not necessarily the case. A levelized cost is nothing else than a mathematical conversion of variable annual costs into a stream of constant payments, with the same present value as the total cost incurred by a power plant over its operating life. The result, in the case of electricity producing projects, is a cost per unit of electricity produced. This cost is equal to the minimum average price at which electricity must be sold to reach breakeven, which means that all studies presenting such production cost figures by definition presents levelized costs. Taxes and subsidies can be accounted for in the same way if so requested. The differences in basic electricity production cost methodologies mentioned so far are important examples of how studies computing production costs may differ. Some studies use simple formulas, others complex models. In some cases only cost figures and assumptions are presented but not the methodology itself. However, several other differences and issues with respect to production costs methodologies can be mentioned. A notable example is the assumption on discount rates, a key factor in determining electricity production costs (see Equation 1). Discounting is used to determine the present value of a stream of future cash flows and it has a significant impact on electricity production costs, since the choice of the discount rate is one of the major cost factors for capital-intensive power generating technologies. In this regard, the reviewed studies differ considerably in their discount rate assumptions, making the stated cost figures not only difficult to compare, but also to generalize as the presented production costs are dependent on the used discount rate. Other areas where studies tend to differ include different assumptions for the power plant economic life, the inclusion of decommissioning and refurbishment expenses, the assumptions on residual values, and for the inclusion of flue gas treatment and ash management costs. Regional conditions seem to explain some of the differences in the reported cost figures, meaning that it will be difficult to compare or use costs originating in foreign studies. A key observation by IEA & NEA [5] is that country-specific circumstances, such as market conditions and availability of resources, are among the main determinants when computing electricity production costs. Time and economy-of-scale are other important factors influencing production costs. The IHS CERA Power Capital Cost Indexes have shown considerable increases in both European and North American power-plant capital costs since the year 2000 (Figure 6). They also reveal that cost changes tend to occur rapidly in time. Similar observations can be found in Elforsk [8], showing a compilation of some Swedish and international indexes related to power-plant capital cost increases: all indexes point to significant cost increases during the years 2000 2010. When it comes to production costs, significant cost changes (both positive 11

and negative) are observed in updated studies, such as those in Elforsk [8, 31], and in the latest DECC studies [10, 12, 32]. A strong influence of economies of scale, meaning that production costs are also significantly affected by the installed capacity, i.e. plant size, was highlighted in Elforsk [8] and many other reviewed studies. Figure 6. Time evolution of the Power Capital Cost Index (PCCI) for North America. This index tracks the costs associated with the construction of a portfolio of 30 different power generation plants in North America. Data source: IHS [33] The methodological differences and the different assumptions create an evident risk of comparing apples with oranges when considering the electricity production costs of the different studies. Since all cost factors cannot be discussed in this review, the focus below will be placed on some elements that exhibit significant impact on the stated production costs. 3.2 Investment costs Investment costs include overnight investment costs (see Figure 7), which is the cost incurred for building a power plant immediately, i.e. overnight. It does not include any assumptions on interest expenses that occur during the construction period. Overnight investment costs are often accompanied by assumptions on interest during construction (IDC) expenses. Examples where studies differ include the assumptions on overnight cost increases due to first-of-a-kind problems, on green field or brown field conditions, and on inclusion of gridconnection costs. For example, first-of-a-kind projects are often associated with higher costs than latter projects due to missing learning effects, i.e. they cannot benefit from lessons already made. A green field project is constructed on undeveloped land, whereas brown field alternatives are built on abandoned industrial sites ready for re-use. The overnight cost components are difficult to identify due to lack of transparency, making it difficult to distinguish where studies differ specifically and why they present such different investment cost figures. 12

USD/kWe 13000 12000 11000 10000 9000 8000 7000 6000 5000 4000 3000 2000 1000 0 upper quartile box lower quartile box mean Figure 7. Overnight investment costs presented in the reviewed studies (i.e. [4], [5], [7], [8], [10 15], [20], [36]). The cost figures are visualized in box plots showing the minimum value, lower quartile, median, upper quartile, and the highest value. The mean value is also included in the figure. Observe that the presented overnight investment costs are expressed per unit of installed capacity, whereas other cost factors are expressed per unit of electricity produced, i.e. levelized costs. The most common way of accounting for the total investment costs is to add the assumed interest expenses (occurring during the construction period) to the assumed overnight investment cost. Interest expenses are dependent on construction period, construction period interest rate, cost outlay, and overnight cost, which all differ depending on the study examined (see [29] for details). Such differences might have a non-negligible impact on production costs, especially when considering technologies associated with long construction times and high capital costs. Some studies [7, 13] do not mention the IDC costs. This is an issue, since if the IDC costs are not included; the cost difference between overnight and total investment costs (including IDC construction costs) might be substantial. One explanation for the wide range of estimates reported for the overnight investment costs of CHP-B, PV, and hydropower is economies of scale, where the overnight investment cost per installed unit of installed capacity is generally smaller for plants with larger power output. Moreover, investment costs for hydropower plants also tend to be highly site specific [34]. Similar observations can be made for PV, where the type of application, e.g. rooftop versus open-air installation, is a major cost factor [20]. 3.3 Fuel and O&M costs Fuel costs are heavily affected by the assumptions on the fuel price and plant efficiency and they both have to be considered when looking at the presented fuel cost figures (see Figure 8). 13

USD/MWhe 170 160 150 140 130 120 110 100 90 80 70 60 50 40 30 20 10 0 upper quartile box lower quartile box mean CCGT Hard coal Lignite Nuclear fuel CHP-B CHP-CCGT Figure 8. Levelised fuel costs presented in the reviewed studies (i.e. [4], [5], [7], [8], [10], [12], and [14]). The cost figures are visualized in box plots showing the minimum value, lower quartile, median, upper quartile, and the highest value. The mean value is also included in the figure. Note that CHP-W fuel costs are omitted, since the costs tend to be negative due to gate fees. Reported fuel cost figures for CHP-W is in the range between -98 USD/MWhe and -246 USD/MWhe. Fuel prices are treated quite differently in the reviewed studies. IEA & NEA [5] and Elforsk [8] assume fuel prices to be constant in time. CASES [14], EUSUSTEL [35] and NEEDS [4] assume increasing prices for some fuels, and oscillating or constant prices for others. DECC [36] and EIA [37] also seem to assume oscillating fuel prices, but actual data can only be found for fossil fuels. The approach used by VGB [7] is unknown, since no fuel price data or other information is provided in their report. Large differences also exist when it comes to the assumed absolute numbers, and in case of CHP-B power plants, also for the choice of the fuel used. Fuel costs for CHP-W plants are negative, since such stations are granted so called gate fees for accepting and processing waste from other parties. Gate fees tend to be site- or country-specific and they are crucial to CHP-W electricity production costs. Assumed power plant efficiencies also affect total fuel costs. A lower electrical efficiency leads to more fuel being spent per unit of electricity produced. The total thermal efficiency of a CHP plant is also very important, as it affects the amount of produced heat. For some power generating technologies, the assumed efficiency figures can be quite different (see Table 4). It is much more difficult to evaluate the reported O&M costs, mostly due to the lack of transparency in the reviewed studies. However, it is important to be aware of the large range of estimates reported for some of the power generating technologies (see Figure 9). Note that EUSUSTEL [15] reports unusually high O&M costs for hydropower, but the reason for such diverging figures from the average estimates are not discussed. 14

Table 4. Reported power-plant efficiencies. Electrical efficiency is defined as the ratio between useful electricity output at a specific time, and the energy value of the supplied energy source during the same time period. The thermal efficiency is the ratio between total energy output, i.e. electricity and heat, and the heat-content of the consumed fuel. Note that the stated efficiency figures over 100% for some CHP plants is due to assumptions on fluegas condensation, making it possible to utilize otherwise lost heat. Such heat energy is excluded when using assumptions on fuel energy content, based on lower heating values. EIA figures are not included because they use higher heating values, making their stated plant efficiencies not comparable with the other studies. Technology Electrical efficiency [%] Thermal efficiency [%] Hard coal 45.0-48-0 Lignite 40.0-45.5 CCGT-C 55.0-60.0 OCGT 38.0-45.0 Nuclear 33.0-36.0 CCGT-CHP 45.0-50.5 89-0-92.0 CHP-B 18.5-30.9 65.0-108.0 CHP-W 19.2-19.4 99.0 USD/MWhe 140 130 120 110 100 90 80 70 60 50 40 30 20 10 0 upper quartile lower quartile mean EUSUSTEL maximum hydropower figures 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 100 0 USD/MWhe Figure 9. Presented O&M costs in the reviewed studies (i.e. [4], [5], [7], [8], [10-15], and [18]). Note that two y-axis scales are used, where the second rightmost one is used to visualize the stated maximum EUSUSTEL hydropower figures. 3.4 Heat crediting in combined heat and power plants Combined heat and power plants (CHP) produce both electrical energy and heat energy. Costs borne by a CHP plant could therefore be imputed to both the produced electricity and to the produced heat. A generally accepted methodology in all of the reviewed studies is to modify 15

the CHP electricity production costs by subtracting an assumed value for the heat energy produced, called heat credit. CASES [14] assumes that the produced heat replaces heat output from an alternative natural gas boiler plant. The heat credit should therefore be equivalent to the cost of fuel that otherwise would have been spent in the gas boiler. Elforsk [8] uses multiple heat credit assumptions and states that regardless of the site-specific situation, a heat credit equivalent to variable district heating costs can always be motivated. Such heat credits are computed on the basis of the cost of fuel usage, as well as variable O&M costs in an alternative biomass-fired boiler plant. Fixed heat credits based on fixed O&M costs and the investment cost for an alternative heat-only boiler plant are also used in the study. DECC (see [10] for details) bases its heat credits on avoided gas boiler costs. However, lack of transparency makes it difficult to understand its methodological approach. Avoided capital and O&M costs are based on a heat incentive by DECC [36]; whereas avoided fuel and carbon dioxide emission costs are based on DECC gas and carbon price projections. However, the heat incentive actually seems to be a fixed tariff rate [38], and where electricity production costs are presented, like in Arup [10], heat revenues appear as a fixed negative operating cost. This makes it difficult to understand how heat credits actually are accounted for. IEA & NEA [5] use a very simple approach, assuming a fixed heat credit per MWh of heat energy produced. EUSUSTEL does not mention the issue of heat credits in their main technical report [15]. However, heat credits appear in their cost calculation appendixes [39, 40], but without a reference or explanation. It is evident that various studies treat heat credits quite differently. A great deal of caution is therefore recommended when looking at CHP production cost figures. The proportion of heat energy production in CHP plants tends to increase with lower plant electrical efficiency, which raises the significance of heat credits [5]. 3.5 Assumed electrical energy production One factor that is extremely important for the estimations of production cost is the amount of electricity that is assumed to be produced, since the costs are generally expressed per kwh of produced electricity. This means that, for instance, a doubling of assumed production will cut the cost per kwh in half. The capacity factor is a common way of theoretically describing the output of a power plant by expressing the assumed production as a fraction of the rated power during a certain period of time. It is defined as the ratio of actual energy produced to the theoretical maximum of running constantly at the installed power capacity. It should be added that the capacity factor does not express when in time the electricity is produced, but only the fraction of the theoretically maximum production that is actually produced, or assumed to be produced. Power generating technologies such as nuclear, coal and CCGT are mostly assumed to run in baseload 5 mode with capacity factors close to 0.85 (see Table 5). However, for both coal and CCGT, the span is quite large. OCGT is treated differently with capacity factors ranging from 0.18 to 0.86, implying that it covers the entire operating span, from pure peak load to base load operation. 5 Baseload power plants meet a region s continuous energy demand. They typically run at all times except in the case of repairs or scheduled maintenance. Peaking power plants usually run when there is a peak in demand for electricity, e.g. during evening hours when household appliances are heavily used, or during cold winter days and hot summer days. 16

Table 5. Assumed capacity factors for non-renewable power generating technologies. Study Capacity factor NP CC CCGT-C CCGT-CHP OCGT IEA & NEA [5], min 0.85 0.85 0.85 0.85 - IEA & NEA [5], max 0.85 0.85 0.85 0.85 - IEA & NEA [5], med 0.85 0.85 0.85 - - CASES [14] 0.86 0.86 0.86 0.86 0.86 EUSUSTEL [15] 0.90 0.85 0.90 0.85 - Elforsk [8] 0.87 0.78 0.78 0.56 - SKGS [13] 0.90 - - - - DECC [9] 0.87 0.95 0.93 0.93 0.18 EWEA [18] - - - - - NEEDS [4] 0.84-0.82-0.57 EIA [11] 0.90 0.85 0.87-0.30 VGB [7] 0.90 0.86 0.68-0.68 An interesting observation can be made for CHP technologies (Table 5, Table 6). While most of the studies assume capacity factors high enough to imply year-round operation, Elforsk [8] assumes that a CHP plant operates during the heating season only. Therefore, its assumed capacity factor is significantly lower than those reported in the other reviewed studies. An exception is made for CHP-W, because such stations have to be used in baseload operation due to the high investment costs and demand all year-round for accepting waste. The reason why other studies assume all year-round CHP operation is unknown. The different assumptions on capacity factors presented in this section are high enough to cause notable differences in estimated production cost. Moreover, production costs are also more sensitive to incremental changes in the assumed capacity factor. An assumed capacity factor of 0.3 versus 0.2, has a significantly larger impact on production costs compared to that of 0.9 versus 0.8 (see [29] for details). Capacity factors for intermittent technologies also tend to be highly site-specific, why expressing production costs of electricity as one fixed number for such technologies is questionable. Table 6. Assumed capacity factors for renewable power-generating technologies. Study Capacity factor HP, Reservoir HP, ROR CHP-B CHP-W WP on WP off PV IEA & NEA [5], min - - - - 0.25 0.37 0.10 IEA & NEA [5], max - - - - 0.23 0.37 0.23 IEA & NEA [5], med - - - - 0.26-0.13 CASES [14] 0.34 0.57 0.86-0.30 0.46 0.12 EUSUSTEL [15] 0.80-0.91 0.57-0.8 - - 0.23-0.29 0.29-0.5 0.15 Elforsk [8] 0.46-0.55 0.76 0.3-0.35 0.34-0.35 0.12 SKGS [13] 0.46 - - - 0.26 - - DECC [9] 0.46-0.77 0.83 0.25-0.29 0.38 0.11 EWEA [18] - - - - 0.17-0.33 0.41-0.42 - NEEDS [4] - - 0.91 - - 0.46 - EIA [11] 0.52 - - - 0.34 0.34 0.25 VGB [7] - 0.80 - - 0.21 0.37-0.43 0.23 17

3.6 Carbon dioxide emission costs Some studies account for carbon dioxide emission fees which penalize fossil-fuelled power plants. Such costs could be seen as external costs, but some studies choose to internalize direct carbon dioxide emission costs, i.e. emissions occurring during power plant operation, into presented electricity production costs, why we opted to keep them in this paper. Discrepancies in the reported levelized carbon dioxide emission costs are due not only to different assumptions for carbon dioxide emission fees (Table 7), but also to the assumed power plant efficiencies (Table 4) and to the fuel carbon content. However, none of the reviewed studies provide any information about the relative fuel carbon content and how they calculate the carbon dioxide emissions. It is therefore difficult to completely understand how these studies arrive at such different results. Table 7. Assumed carbon-dioxide emission fees in the reviewed studies -if reported- [USD/metric ton CO 2 ]. Study Fee VGB [7] 41.8 DECC [9] 21.9 IEA & NEA [5] 31.2 EUSUSTEL [15] 15.3 It is important to be aware of whether or not the carbon dioxide emission costs have been included, as they might have a significant impact on the electricity production costs (see Table 8). For instance, the CO 2 emission costs for hard coal used by DECC [12] is high enough to completely change the picture of production costs for coal power compared to other electricity generating technologies. Table 8. Reported levelised carbon-dioxide emission costs [USD/MWh e ]. Study CC hard coal CC lignite CCGT-C CCGT-CHP OCGT VGB [7] 37.6 39.0 19.4-22.2 DECC [9] 78.1-29.6 - - IEA & NEA [5], median case 24.9-10.9 - - IEA & NEA [5], min 22.9 26.2 10.6 9.4 - IEA & NEA [5], max 24.5 28.3 11.7 15.5 - EUSUSTEL [15] 10.6 13.6 5.2-5.6 - - 3.7 Backup costs for intermittent technologies Some studies (i.e. [14], [15], [41]) argue that intermittent, non-dispatchable, power generating technologies should include the backup costs due to their fluctuating electrical energy production. Backup power is seen as a necessity because of the inflexible, variable and also unpredictable behavior of the intermittent technologies. Backup costs should be seen as external costs, as they are shared amongst all producers, and hence society, and not directly accounted for by individual producers. We opted to keep them as some studies internalize them into stated electricity production costs. EUSUSTEL accounts for backup costs in their cost calculation appendixes [39, 40] but data and information cannot be found in the main reports [15, 35]. Another transparency issue is that neither EUSUSTEL nor NEEDS describe the methodology used for calculating the backup costs. This makes it impossible to exactly deduce how backup costs are accounted for. Interestingly enough, EUSUSTEL also seems to assume that large-scale reservoir hydro should be subject to backup costs [39, 40], but no justification for this assumption is provided. 18

Backup power is assumed by CASES [14] to be delivered by either a hard coal or a CCGT power plant, and the same assumptions are made in the EUSUSTEL and the NEEDS studies. Figure 10 shows the backup cost formula used by CASES [14]: C BU = A h k v Ak P h w = A k 1 hf P h W C BU h h v w P A k = Cost Back-Up = Full loading hours, supply = Full loading hours of the renewable power plant = Power credit of the renewable energy plant = Annuity, incl. the annual fixed costs for the Back-Up technology Figure 10. Backup cost formula as used by CASES [42]. Even though CASES provides the formula used, it is still difficult to understand exactly how the backup costs are calculated. No calculation examples are provided and the terms included in the backup cost equation (Figure 10) are not properly explained. For example, it is unclear which type of annuity is used; the annuity-due or the annuity-immediate, and the inputs required for its computation are not provided. The issue of backup costs introduces a great amount of uncertainty, both with regard to their computation and whether they should be accounted for or not. It is also questionable if they should be internalized into the electricity production costs as they are likely not paid for by individual producers. However, backup costs and other electricity system cost factors (see chapter 4) are perhaps as important in decision and policy making as other external cost factors, so that they should be included when discussing total electricity generation costs. The assumptions about such costs are by no means negligible. The inclusion of backup costs for wind power causes production costs to increase by 15-50% in reviewed studies, depending on the backup technology assumed (see [29] for details). As such, backup costs can constitute a major part of total wind power electricity production costs, thereby complicating the comparison with production costs presented in other studies. Please note that other intermittent power sources, such as photovoltaics, also are in need of backup power and are subject to backup costs. 3.8 Issues related to transparency and input data A general problem with the reviewed studies is the lack of transparency, making it difficult to understand and analyze the different steps in computing the electricity production costs. This is also the main reason why it is difficult to address the impact of certain factors on the reported results. Lack of transparency includes confidential data sources and models, poorly explained formulas, lack of concrete calculation examples, and unknown or unexplained assumptions and approaches. These points complicate the understanding of the methodology used in the cost calculations, thereby causing credibility and suitability concerns. When analyzing the presented cost figures, for example investment costs and O&M costs, these are generally presented as a single figure, meaning that the individual components are not specified. Thus, it is difficult to assess where costs differ and the reasons for discrepancies between studies. In most cases, it is unknown whether the presented costs are based on existing power plants, power plants under construction, planned projects, literature data, or something else, since this information is rarely specified in the reviewed studies. Another issue is that some studies tend not to differentiate between power generating technologies with similar, but not equal, technical characteristics. For example, different kinds 19

of solar generating technologies such as crystalline silicon and thin-film solar cells are commonly grouped together and named photovoltaics and nuclear power are for some of the studies referred to as PWR, even though some alternative reactor technologies such as BWR are used. What kind of power plant is actually being assessed can be defined quite differently. The actual power plant referred to can be a single wind turbine or a large wind farm, pulverized coal combustion or advanced pulverized coal combustion, or sometimes not specified at all. Some studies define technical characteristics quite specifically, while in others they are not given much attention. Cost and technical input data are also frequently decoupled, meaning that such data are collected from different sources and no direct resemblance exists. In many cases it is difficult to verify whether the reviewed studies mix different, but similar, power generating technologies (such as different types of photovoltaic technologies, for example) into a common category (such as photovoltaics ). Moreover, the lack of transparency makes it difficult to assess whether generalizing technologies or decoupling data sources have any significant impact on the presented electricity production costs. 4. Policy implications The levelized cost of electricity, or similar definitions (see the previous discussion in Section 3), has become the most popular methodology for comparing electricity production costs from different technologies. The methodology is focused on the producers, but what about the total costs that have to be paid by the consumers or society? Production costs are in most cases limited to the plant-level, meaning that the system border is placed where electrical energy is fed to the grid 6. This implies that the current levelised cost methodologies do not focus on a power plant s different impacts on the electricity system as a whole. As pointed out earlier in this paper, some studies account for backup cost of intermittent power sources (section 3.7) and grid connections costs (section 3.2). However, while grid connection costs are based on standard values which are common for most power generating technologies, other factors such as costs for providing reserve power and long distance transmission investment costs are not considered. Current energy policies for electricity generation aim, among other things, to increase renewable electrical energy production, diversify production to improve energy security, and reduce carbon dioxide emissions due to electricity generation [43]. However, it is important that such policy measures are implemented on a cost-effective basis, because there might be other, more cost-effective ways of reaching the specified targets. While the views on how much more costs would be added to the total production costs can be rather different, there is widespread agreement that cost increases would be more substantial for intermittent, nondispatchable, power sources ([5], [44], [45]). Table 9 shows a compilation of some reported estimates for system integration costs for wind power 7. The fact that current levelised cost methodologies do not incorporate electricity system costs in a consistent way needs consideration if the presented production costs are to be used for planning and policy making. Integration costs of intermittent power sources should not be considered as ways of discriminating such technologies. Instead, they should be considered as costs based on inherent technical aspects to be considered when evaluating electricity generation costs; much like when evaluating emission-related external costs of fossil-fuelled power plants. Moreover, integration of renewable power sources can bring about benefits, 6 By "system border" we mean a boundary that separates the internal components of the power plant from external entities, which are excluded from the scope of the study. 7 The studies presenting these estimates were not thoroughly reviewed with respect to the used methodologies and the underlying assumptions; the estimates presented here are reported only to show some quantitative assessments for system integration costs, as well as the large dispersion that characterizes these estimates. 20